$137M in AI-Generated Political Ads Flood US Midterms as Most Voters Fail to Spot Deepfake Content

AI-Generated Political Ads Dominate 2026 Midterm Elections

The 2026 US midterm elections have witnessed an unprecedented surge in AI-generated political ads, with campaigns and PACs spending more than $137 million on fabricated content1

1

. This represents a massive increase from less than $50,000 in AI ad spending during the entire 2024 election cycle2

2

. The Wesleyan Media Project has identified 220 political ads nationwide created at least partially using AI tools, though researchers acknowledge this figure likely represents an undercount due to the difficulty in definitively determining AI usage3

3

.

AI in elections has evolved from a predicted threat to an operational reality. Eighty-one percent of AI-generated midterm ads were sponsored by Republican candidates or groups, 12 percent by Democrats, and the remainder by independent candidates or groups2

2

. The AI ad blitz has fundamentally altered the information landscape as the November 3 vote approaches, with reality-bending deepfake ads proliferating across television airwaves and social media platforms.

Deceptive AI Ads Target Candidates Without Disclosure

Many AI-generated political advertisements are designed to deceive voters without proper labeling. A particularly concerning example features Texas Senate candidate James Talarico in an 85-second video where he appears to read old tweets aloud and voice commentary he never actually made1

1

. UC Berkeley deepfake expert Hany Farid noted that most people wouldn't immediately recognize it as fabricated content, with the "AI GENERATED" tag appearing small and faint in a corner.

Deceptive AI ads have targeted multiple candidates with fabricated scenarios. Front Line Action, a political action committee aligned with Republican Senator Ted Cruz, spent over $92,000 on an ad opposing Democratic candidate Abdul El-Sayed, depicting him playing a board game with cards reading "Death to America" and "Celebrate 9/11"2

2

. Another Make Oklahoma Great Again PAC ad showed gubernatorial candidate Mike Mazzei alongside Hillary Clinton under the words "Clinton Lover. Trump Hater," with no AI label, spending more than $200,000 to air it on Oklahoma television1

1

.

Voters Cannot Identify Deepfakes Despite Media Literacy

Research reveals a troubling reality: voters correctly identify deepfakes only about 16% of the time1

1

. A June study by Utah Valley University found that people who called themselves "extremely familiar" with deepfakes performed no better than those who claimed no familiarity at all. This demonstrates that media literacy alone doesn't inoculate voters against fabricated content.

The ethical implications of AI become clearer when examining voter perception. A 2021 experiment with 5,750 people showed that a deepfake of Elizabeth Warren convinced 47% that a fabricated scandal was real, while a plain fake headline convinced 43%1

1

. Recent surveys by the nonprofit AAPC Foundation found that 73 percent of voters consider it unacceptable for political campaigns to use AI to generate images or video in ads, while 95 percent find it unacceptable to create videos of opponents saying things they never said3

3

.

Slopaganda and the Spectrum of AI-Generated Content

Researchers have coined the term "slopaganda" to describe AI-generated political content designed to stir emotions rather than present coherent arguments. NYU's Scott Babwah Brennen characterizes this genre as political slop flooding the information ecosystem1

1

. Not all AI-generated political advertisements are equally problematic. Some serve cosmetic purposes, such as animating a still photo of Oklahoma candidate Chip Keating into a smile, or creating synthetic family scenes where campaigns previously used stock footage. Campaign ad veteran Fred Davis told The New York Times that scenes costing $2 million to shoot now cost approximately $200,000 to generate1

1

.

Other ads fall into obvious satire territory, including New York's governor and mayor gardening together while pitching a "political delusion pill," or Spencer Pratt as Batman leading Angelenos armed with tomatoes1

1

. These political cartoons represent protected satire rather than intentional disinformation.

Regulatory Challenges and Ineffective Disclosure Requirements

At least 31 states have enacted deepfake election statutes, yet these laws have proven largely ineffective1

1

. Wesleyan researchers found that only 27 percent of AI-generated midterm ads carried AI disclaimers, with 29% labeled in states with laws compared to 32% in states without legislation1

1

3

3

. No federal mandates currently require political ads to carry AI disclaimers, though some states have legislated restrictions or disclosure requirements2

2

.

Even when AI disclaimers appear, they often fail to inform voters effectively. NYU's Center on Technology Policy showed 1,051 people labeled and unlabeled ads, finding that 37% didn't remember seeing Michigan's required disclaimer at all1

1

. Because the disclaimer stated "manipulated by technical means" rather than explicitly mentioning AI, only 21% identified AI as the technology involved. Democratic Senator Jon Ossoff called the trend "deeply troubling" and warned it represents "a slippery slope toward the destruction of truth in political debate"2

2

.

TrueMedia.org and Deepfake Detection Efforts

AI researcher Oren Etzioni founded the nonprofit TrueMedia.org in 2024 to combat political deepfakes, developing state-of-the-art deepfake detection technology that analyzed more than 60,000 images, videos, and audio clips in elections worldwide1

1

. After open-sourcing the technology and transferring it to Georgetown University, researcher Sejin Paik resurrected TrueMedia.org in time for the 2026 election cycle peak. Etzioni initially expected foreign operatives to be the primary source of fabricated content, but the deepfakes in 2026 are predominantly coming from American political campaigns and PACs on television and social media feeds1

1

.

Impact on Voter Mobilization and Election Integrity

While concerns about election integrity continue to mount, researchers suggest the impact of slopaganda on actual election outcomes may be limited. New York University's Center on Tech Policy concluded that "political slopaganda is likely more effective at mobilizing voters or raising money than it is in impacting candidate selection"3

3

. The report noted that like digital memes, AI-generated content may have limits to its staying power as audiences become increasingly inundated with fabricated material.

The London-based Institute for Strategic Dialogue has characterized the 2026 midterms as the first US election in which AI is built into every major social media platform3

3

. This structural shift in how political campaigns operate raises questions about long-term implications for democratic discourse. Watch for proposed federal legislation requiring plain-language AI labels both on-screen and in audio for paid political ads, as well as potential legal challenges to deceptive AI ads following the November vote.